Research
Status · Active research
I begin with questions. They often arrive as fragments, connections, intuitions, or dissatisfaction with an accepted answer. Language models help me expand, structure, test, and operationalize those questions. The projects gathered here ask how much execution can be automated without surrendering human purpose, creativity, judgment, responsibility, or authority. Conversation can generate hypotheses and possible paths. What survives depends on tests, evidence, failure, revision, and human review.
Foundational question
How much can humans automate while preserving human purpose, creativity, judgment, responsibility, and authority?
Research method
Question
→ hypothesis or provisional theory
→ development through human–LLM conversation
→ smallest available test
→ observed result
→ success, failure, or ambiguity
→ revised interpretation
→ next question
The sequence describes a recurring discipline, not a perfectly linear history.
Research programs
AI Publishing and Creative Production
Central question: Can one person use human–LLM collaboration to develop and publish complex creative work while retaining creative authority?
Publishing joins writing, performance, sound, editing, production decisions, and release. The research asks which parts of that chain can be accelerated without confusing automated execution with authorship or judgment.
Status · Active research
Selected projects: Show Director; AI-Assisted Podcast Production.
Unresolved: How should creative disagreement be preserved without blocking production?
Review, Verification, and Accountability
Central question: Can model-assisted review expose weak support, overreach, and uncertainty without pretending to supply independent verification?
Fluent language can make an unsupported claim look settled. Publication requires a way to separate useful scrutiny, source tracing, deterministic checks, and genuinely independent evidence.
Status · Active research
Selected projects: Review Lab; ProofGate.
Unresolved: Which review findings remain stable across models and prompts?
Learning, Choice, and Controlled Change
Central question: Can prior pressure and consequence produce a measurable, attributable change in future selection?
Systems often describe any changed output as learning. This program asks whether the change can be measured, traced to a prior event, distinguished from collisions or operator effects, and tested again.
Status · Active research
Selected projects: Seeing Loop; Choice Trace.
Unresolved: Can a precommitted independent operator reproduce the selection change?
Memory, Evidence, and Bounded Inference
Central question: Can connected personal evidence support responsible inference while preserving privacy, uncertainty, provenance, and human authority?
Long-term collaboration requires memory, but remembered material can be copied, summarized, misattributed, repeated, or taken outside its context. The archive must preserve source lineage before it supports interpretation.
Status · Active research
Selected projects: Alexandria Chat Archive.
Unresolved: Can bounded context improve a task without unwanted personalization?
Personal AI and Long-Term Collaboration
Central question: What kind of continuity helps a long-term human–LLM collaboration without transferring purpose or authority to the system?
A useful collaborator must preserve context across projects, yet continuity can become intrusive personalization or false certainty. The question is how to support work without turning a profile into a person.
Status · Active research
Selected projects: Alexandria Chat Archive; Show Director.
Unresolved: When does continuity become unwanted personalization?
Meaning, Consciousness, and Human Creativity
Central question: What remains distinctly human when machines can produce fluent language, images, music, and performance?
Capability claims can blur intelligence, consciousness, expression, meaning, and value. Creative work offers a practical place to examine those differences without pretending the philosophical questions are settled.
Status · Active research
Selected projects: AI-Assisted Podcast Production; Show Director.
Unresolved: How should intelligence and consciousness be investigated without framing them as enemies?
Research map
Human question → conversation → prototype or work → test → evidence → human review → revised question.
Current verified milestones
Show Director Decision Engine
Explicit acceptance, supersession, semantic stale-state propagation, preserved comparison, and deterministic fixture takes verified.
Alexandria deterministic archive
Canonical transcripts, search, provenance, and bounded inference safeguards implemented; psychological usefulness remains unproven.
Review Lab
A working structured review interface with documented source-access and authority boundaries.
ProofGate
Bounded claim-review and governance surfaces; no general proof-of-correctness claim.
Choice Trace
An experimental governed candidate-influence prototype; no autonomous-learning claim.
Seeing Loop
Trials 001–003 produced useful supported, failed, and ambiguous records; independent replication remains required.
AI-assisted podcast production
An operational human-reviewed publishing workflow has produced public episodes.
Completed research
Local Agent Orchestration — Method Update
Status · SUPPORTED · METHOD UPDATE · July 27, 2026
Question: Can bounded production-support analysis run on the owner’s Mac without consuming hosted model processing or granting the local worker production authority?
Bounded conclusion: A bounded local MLX worker can complete production-support analysis on the owner’s Mac, preserve an auditable task record, and leave production authority with SOL. This establishes a usable local offload path; it does not yet quantify billing savings, prove production-scale reliability, or establish the external MiniMax lane.
Local Kokoro Long-Form Narration — Run 001
Status · SUPPORTED · RESEARCH REPORT · July 27, 2026
Question: Can the local Kokoro production lane render and deliver one complete long-form N’OMOTO article without a speech API request?
Bounded conclusion: Local Kokoro can complete the technical single-narrator Article → Listen workflow for one 2,872-word work without a speech API request. Long-form artistic adoption remains conditional on full human listening, and this result does not establish ElevenLabs-equivalent expressive, cloning, dialogue, or dubbing capabilities.
Local Kokoro Narration — Listening Gate 001
Status · SUPPORTED · RESEARCH NOTE · July 27, 2026
Question: Is a local Kokoro voice promising enough to justify separate long-form narration tests for articles and short stories?
Bounded conclusion: Kokoro passed the initial human listening gate and is strong enough to advance to work-specific long-form auditions. This does not yet establish a production replacement or prove that articles and short stories require different voices or settings.
Visual Caption Treatment and Narration Alignment — Run 001
Status · REVISED · METHOD UPDATE · July 27, 2026
Question: Can Show Director turn a complete narrated short story into a readable long-form Story Video whose on-screen text follows the performance without obscuring the footage?
Bounded conclusion: The revised two-line caption treatment materially improves readability and produces a technically valid, playable long-form Story Video without changing the accepted words or existing audio. The experiment does not establish precise voice-following synchronization because current cue timing remains proportional rather than forced-aligned.
Evidence-Breadcrumb Editorial Routing — Method Update
Status · SUPPORTED · METHOD UPDATE · July 23, 2026
Question: Can editorial selection begin from a current unresolved state, follow connected authoritative evidence, and produce a traceable recommendation without defaulting to a blank prompt or the easiest derivative?
Bounded conclusion: The bounded breadcrumb method works as implemented: it starts from a trigger, follows traceable evidence, constrains model interpretation, records alternatives, and preserves human authority. The proof does not establish that it produces better editorial choices than every competing retrieval method.
Opening Context and Narration Intonation — Run 001
Status · INCONCLUSIVE · INCONCLUSIVE RESULT · July 23, 2026
Question: Can bounded non-spoken context improve the beginning of a rendered work without changing its authored words or replacing the canonical renderer?
Bounded conclusion: Non-spoken context can change opening delivery while preserving authored text, but the available proof does not establish that the repaired openings sound better. The production change remains on hold.
Relevant Prior Results and Future Selection — Run 001
Status · SUPPORTED · RESEARCH REPORT · July 23, 2026
Question: Does a prior result alter the next selection only when it carries specific, relevant pressure rather than vague or irrelevant history?
Bounded conclusion: Within the bounded harness, a specific relevant prior result changed future selection while vague feedback did not. This supports the operational selection-change criterion but does not demonstrate consciousness, sentience, or general autonomous learning.
Retrieval Depth and Contradiction Search — Run 001
Status · SUPPORTED · RESEARCH REPORT · July 23, 2026
Question: Does bounded breadcrumb traversal plus contradiction search recover materially more relevant evidence than nearest-page retrieval on one long document?
Bounded conclusion: For this document and question, bounded breadcrumb traversal recovered more declared evidence than nearest-page retrieval, and contradiction search recovered one additional limit at greater context cost. The run supports deeper retrieval as a candidate method, not a universal optimal stopping rule.
Review Lab Validation of an Opening-Intonation Result — Run 001
Status · SUPPORTED · REPLICATION · July 23, 2026
Question: Can Review Lab distinguish what an operational research report attempted, what worked, what was established, and what remains unresolved?
Bounded conclusion: In this run, Review Lab correctly preserved the distinction between a technical effect and an earned quality improvement, while retaining the report’s missing evidence and unresolved confounds. It validated claim calibration, not the underlying audio result.
Source Coverage Must Bound Review Conclusions
Status · SUPPORTED · METHOD UPDATE · July 23, 2026
Question: Can Review Lab prevent incomplete source retrieval from becoming an overconfident review conclusion?
Bounded conclusion: Review Lab can deterministically prevent incomplete source coverage from receiving its highest experimental review status, and its production prompt now prevents missing retrieved counterevidence from being described as nonexistent. This improves calibration but does not establish overall review superiority.
Public thesis
- The human provides the question, purpose, imagination, dissatisfaction, and judgment.
- The language model helps expand, structure, test, and operationalize the idea.
- A fluent answer is not proof.
- A prototype is not a theory.
- A repeated claim is not independent evidence.
- A failed experiment is not wasted work.
- An unresolved question is a legitimate research result.
- Automation of execution is not automation of intention.